The increasing complexity of distributed applications, software services that can be dynamically deployed, adjusted and composed, paves the way for new challenges in software and service engineering. This paper describes a novel approach that combines the exibility of MDE techniques to deal with the conceptual modelling of MAS and the expressive power of OWL based ontologies to deal with semantics constraints veri cation as well as domain knowledge provision of MAS models. We will illustrate these ideas through the modeling of a crisis management scenario, using a rst prototype of our future Design tool: OperettA.
Nowadays' distributed applications based on the notion of service-oriented
systems {which can dynamically adapt, organise, and compose to satisfy with their
networked stakeholders' needs{ are fostering the software engineering research
area with new challenges. As these distributed systems have to be deployed
within real organisational contexts, adhering with organisational rules, and
meeting stakeholders' expectations, it is crucial to characterise the software
architectural and functional requirements in terms of their correlations with the actual
environment. To deliver on this aim, a promising approach in software
engineering has been to build methodologies along with conceptual modelling languages
that better re ect and describe the complex social and human organisational
context where the system-to-be has to be deployed [
In this paper, we describe some achievements and future improvements of
a MAS development framework, OperettA [
Model Driven Engineering (MDE) refers to the systematic use of models as
primary artefacts throughout the Software Engineering lifecycle. The de ning
characteristic of MDE [
The paper is organised as follows: Section 2 presents the methodological context of our approach with a motivation example. Section 3 provides an overview of our approach. Section 4 summarises the main characteristics and bene ts of the approach adopted, which is partially implemented within the OperettA tool. 2 2.1
OperA [
The development framework for agent societies, proposed in OperA, is
composed of three conceptual design models: Organisational Model, Social Model,
and Interaction Model, as detailed in [
Using an example taken from the Dutch procedures for crisis management, we
provide a conceptual model based on organisational and social concepts. The
modelling phase is conducted according to the OperA methodology [
The structure diagram depicted in Figure 1 represents the crisis situation. It
speci es the responsibilities and goals of each role, e.g., each time an emergency
call occurs to the Emergency Call Center, this role alerts the Fire Station entity,
informing about the location in which the (possible) disaster is taking place. The
Fire Station is responsible to build the appropriate re brigade and depends on
the established Fire ghter Team in order to achieve the objective extinguish the
re. When the team arrives at the accident location, it has to decide (based on
its personal experience) the severity of the disaster. Only after this evaluation
is reported, an intervention decision is taken. For example, according to local
rules, the evaluation should comply with some standardised emergency levels,
as established by the Dutch Ministry of Internal A airs. For the sake of
simplicity, we consider that Fire ghter Team sets up a strategic intervention based
on the results of two evaluation criteria: damage evaluation and re evaluation.
Based on the number of wounded, Fire ghter Team decides on the necessity or
not to ask for ambulance service. Moreover, the Fire ghter Team checks if the
damage involves building structures in which case police intervention is
necessary to deviate tra c. From the re evaluation criterion, Fire ghter Team can
decide whether it is the case or not to ask Fire Station for a Fire ghting Truck
intervention. As described in [
First_Aid_ Station (fas)
(as) Police_Station (ps)
Firefighter_
Team (ft) (dt) (dbf)
Fire_Station
(fs) (ef)
(el) Firefighting _Truck (fft)
Emergency _Cal _Center (ecc)
Notice that, the speci cation provided at this time is not su cient to give a complete picture about the know how required to software systems to achieve the modelled organisational objectives. Nevertheless, the level of abstraction achieved provides enough anchor points for agents to coordinate their activity without fully pre-specifying the capabilities of the agents and therefore limiting exibility.
In order to e ectively deal with the MAS development, the proposed approach takes into account both the syntax and the semantics of a MAS, through an integration of MDA with Reasoning and Domain Knowledge speci cation Based on Ontology. Fig. 2 illustrates the architecture of our approach. The central part corresponds to the OperA metamodel, which provides the syntax of MAS. The right part provides the actual semantics of MAS, which is described by the OperA ontology. The MAS ontology instantiation will be automatically produced from the MAS model, which is created with the OperettA tool. Next, the MAS ontology instantiation will be semantically checked against the OperA ontology, see section 3.2. The left part provides an interaction between existing domain ontologies and MAS models. The interaction is maintained by de ning transformations relation between the OperA metamodel and EODM3. EODM is an implementation of the ODM standard from OMG, de ning metamodels for RDF(S) and OWL, see section 3.3.
Metamodel o t s m fr o n o C
Import/export o t s m fr o n o
C
MAS Model (e.g.: FireStation) o t s m fr o n o C
The rst aspect of our integrated approach is directed toward reasoning on our models using logics, with description logic as our language mainly dedicated at reasoning on the structural aspect of our models. This will provide us with the ability to use the power of descriptive logic along with associated reasoners, combined with techniques to formally analyse our models and enhance their quality.
At the meta-level, the abstract syntax is de ned through metamodelling and the semantics is based on description logic. This results in the OperA ontology, which formalises OperA conceptual framework concepts and their relationships, as well as domain independent OperA semantics constraints. Meanwhile, the integration of the metamodelling and ontologies supports domain speci c languages and also o ers the opportunity to query the models. Indeed, as the OperA metamodel is associated to the OperA ontology, a MAS model is associated to a MAS ontology instantiation. Consequently, the semantics of our models are stored in the MAS ontology instantiation, which is automatically produced using 3 http://www.eclipse.org/modeling/mdt/?project=eodm the MAS model (following a straightforward transformation). Hence, it allows for the designer, the veri cation and validation of models using ontologies; namely, the application of description logics for reasoning about the OperA language. In the following, some examples of semantics constraints, that we are interested in verifying, have been proposed.
{ Checking if the objective of a dependency is an objective or a sub-objective of the dependency's initiator. For instance, in the running scenario, this would mean verifying that the role re-station possesses an objective or subobjective extinguish re as it appears to be dependent on Fire ghter Team for it. { Checking if each dependency is realised by at least one scene. { Checking if roles are involved in a dependency, do indeed cooperate in at least one realisation scene of that dependency. Again, referring to the scenario, there must be at least one realisation scene of objective deal with big res dependency in which Fire ghter Team and Fire ghter Truck must cooperate. { Checking that each role posses at least one dependency link.
The approach described above has been partly implemented in and illustrated
by the OperettA prototype [
The second aspect of our approach permits the use of ontology within our models as the source of domain speci c related knowledge, necessary for the description and support of roles interaction and communication in OperA organisation and at a lesser extend, domain speci c semantics constraints enrichment. That is, domain ontologies are integrated at the model level. They are de ned, used and imported within the OperA modelling language as well as exported from it. For the latter two, a relation is de ned at the meta-level, between the Eclipse Ontology De nition Metamodel and the OperA metamodel enabling the interaction with standard ontology representation languages like OWL. Meanwhile, in the ontological world the integration is at the same level. That is, the OperA ontology and domain ontology are at the same level within logic hierarchy; paving the way for the analysis of the overall structural aspect of the organisation, consisting in querying one knowledge base being the combination of the OperA ontology, the Domain Ontology and the derived MAS ontology instantiation.
Firstly, a vocabulary is available for describing interactions and supporting communications related to a domain. Referring back to our Fire Station organisation, given the equivalent formal notation for the objective extinguish {e.g., Extinguish-Fire(L : location){ of the Fire gther Team role, the concept of Location must be de ned in the Domain ontology, otherwise the parameter type will not be available when de ning the objective. Furthermore, at run-time, this ontology will be used by members of the organisation in order to communicate.
Secondly, the OperA Ontology can be enlarged by further modelling domainspeci c constraints enriching the generic semantic constraints. It provides the opportunity to de ne modelling rules within the domain ontology as (re)con guration rules e.g., a speci c domain could forbid more than 2 dependencies between two speci c roles.
This second aspect provides a separation of concerns for knowledge taking into account its domain speci c part and enabling at the same time the tool to tailor himself based on it. 4
This paper describes the features of an implemented design tool, OperettA, based on the OperA methodology. This enables the construction of a development framework to support software and services engineering. Besides, this contributes to the achievements of a more general research objective established within the ALIVE project. Speci cally, the ALIVE project combines cutting edge Coordination technology and Organisation theory mechanisms to provide exible, high-level means to model the structure of inter-actions between services in the environment.
The proposed approach (partly) implemented in OperettA deals with techniques to integrate features of the model driven architecture (MDA) with features of the ontology languages (OWL). Main bene ts of our approach come from the fact that it provides a model driven approach for the speci cation of a MAS dealing with its semantics and its provision of domain speci c knowledge. Hence, our approach allows designers for powerful reasoning mechanisms to be employed as well as a smart integration of domain speci c knowledge that comes rst to re ne and enrich (along with further constraints) the speci cation of the design model.